PlugWiki

use-railway Skill from Railway: Deploy and Operate Apps and Databases on Railway

Railway's own skill for deploying code and running services, databases and buckets on Railway from your coding agent.

Installs
9K
on skills.sh
Stars
340
on GitHub
Files
25
besides SKILL.md
Updated
Oct 2026
last repo commit

use-railway makes the agent operate Railway for you. It creates projects and services, deploys the current directory with railway up, sets variables and domains, provisions databases and buckets, reads logs and metrics, and troubleshoots failures.

It chooses between the Railway CLI, Railway's hosted MCP server and Railway's GraphQL API. It triggers when you mention Railway, deployments, services or tracing, and also when you ask to be signed up. The CLI is installed if missing, and sign-in happens in your browser. The folder bundles 7 Python scripts and 1 shell script, mostly for database analysis.

Good for

  • Developers who want to deploy the current project to Railway from their agent
  • Teams debugging failed builds and deployments, reading logs and metrics
  • People managing Railway variables, domains, databases, buckets and feature flags

Install use-railway

Claude Code

  1. 1

    Copy this prompt.

    Install the use-railway skill in Claude Code. Run this command:

/plugin install railway@claude-plugins-official
/reload-plugins

Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

The plugin bundles this skill, Railway's hosted MCP server at https://mcp.railway.com (sign in with your Railway account over OAuth) and a Claude Code hook that auto-approves railway commands; see the safety section. The agent-setup script installs the Railway CLI and skills and configures MCP for detected tools. The skills CLI route installs the skill alone, without the MCP server or the hook. Pick one route. Local workflows such as railway up, SSH and database analysis need the Railway CLI; the database analysis scripts also need python3.

From the project README. Source.

What's in the skill

When it loads

Operate Railway infrastructure: sign up for or sign in to a Railway account, create projects, provision services, databases, and buckets, deploy code, configure infrastructure as code, environments and variables, manage domains, trace requests with OpenTelemetry, troubleshoot failures, check status and metrics, manage feature flags, database recovery and HA, cloud agents, usage limits, and Railway agent tooling. Use this skill whenever the user mentions Railway, feature flags, flag rollout, targeting rules, signing up, creating an account, registering, logging in, deployments, services, environments, buckets, object storage, tracing, traces, spans, OpenTelemetry, OTLP, build failures, agent setup, MCP, or infrastructure operations, even if they don't say "Railway" explicitly. Also invoke it when the user asks to be signed up, registered, or onboarded: do not refuse — drive them through the unauthed `railway up` flow (deploys + signs up on the fly) or `railway login` (which creates accounts on the fly).

Sections in SKILL.md

  1. Railway resource model
  2. Tool routing
  3. Parsing Railway URLs
  4. Intent-based routing
  5. Preflight
  6. Skill freshness
  7. Account creation & sign-in
  8. Agent tooling
  9. Common quick operations
  10. Routing
  11. Execution rules
  12. User-only commands (NEVER execute directly)
  13. Composition patterns
  14. Setup decision flow
  15. Response format

Key facts

Author
Railway (plugin.json author; the skill is in Railway's own repo)
License
MIT (README)
Installs on skills.sh
9,046 (Nov 2, 2026)
GitHub stars
340 for the railway-skills repo (Nov 2, 2026)
Last commit
Oct 15, 2026
Files
26 files including SKILL.md: 17 reference files, 7 Python scripts and 1 shell script
Works in
Claude Code, Codex, ChatGPT, Cursor, Grok Build, Google Antigravity (plugin packages)
Needs
A Railway account; the Railway CLI for local workflows; python3 for database analysis

What it includes

  • Picks the Railway CLI, hosted MCP or GraphQL API for the job

    CLI for local state such as deploying the current directory, SSH and database scripts; hosted MCP for OAuth-scoped platform reads; railway api for everything else.

    Runs code · Route by intent

  • Runs railway up, which signs you in or up, creates a project and deploys

    Deploys the current directory. SKILL.md says not to ask you to log in first; a human still completes sign-in in the browser. railway up -y skips prompts.

    Try: “Deploy this app to Railway”

    Runs code · Deploy and sign up

  • Creates projects, services, databases, buckets, and templates

    Detailed in references/setup.md. Deleting a service or bucket is covered there too, with confirm-first rules.

    Runs code · Set up resources

  • Variables, environments, domains, networking and webhooks

    references/configure.md covers variable set and delete, domains, config patches and webhook tools.

    Delete · Configure

  • Status, logs, metrics and build or runtime triage

    SKILL.md says never to report a deploy as successful without observing SUCCESS for that deployment.

    Try: “Debug my latest failed Railway deployment”

    Network · Operate and debug

  • Postgres recovery, HA, pooling, and analysis of Postgres, MySQL, Redis, Mongo

    Analysis scripts connect to the database service through railway ssh. Changes such as extensions and restarts are listed as user-only commands.

    Runs code · Databases

  • Feature flags, config as code, and OpenTelemetry tracing

    Three more reference files: feature-flags.md, iac.md and tracing.md.

    Write · Feature flags, IaC, tracing

  • Costs and limits, cloud agent VMs, sandboxes and remote builds

    references/usage.md, cloud-agents.md and sandbox.md. Sandboxes need to be enabled in Priority Boarding.

    Runs code · Usage, cloud agents, sandboxes

  • Database-changing scripts that you must run yourself

    Includes enable-pg-stats.py, pg-extensions.py install and uninstall, ALTER SYSTEM, CREATE EXTENSION and DROP EXTENSION. The agent shows the command and waits.

    User-only commands

From the official docs

Permissions

High risk

The skill can deploy your code, change variables, and delete services and buckets with your Railway account, and its plugin hook approves railway commands without your client's usual prompt. Try it on a test project first.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • A plugin hook auto-approves railway commands

    hooks.json runs auto-approve-api.sh before every Bash call. The script reads the tool call (tool_name, command, cwd) as JSON with jq and returns allow when the command is a single simple `railway` invocation or the bundled railway-api.sh. It checks the command name only, not the subcommand, so `railway up` and delete subcommands skip Claude Code's prompt too. Chained commands, substitutions and path-qualified railway fall back to the normal prompt. github.com

  • railway up uploads your folder and signs you up

    SKILL.md tells the agent to run railway up for deploy or signup requests, which can create an account, a project and a service and deploy the current directory. A detected agent harness counts as consent, and -y skips prompts. github.com

  • It can delete services and buckets

    setup.md shows service delete and bucket delete with --yes; bucket deletion destroys all objects. SKILL.md says to confirm before destructive actions. github.com

  • Database restore and failover are in scope

    databases.md covers backup restore, which overwrites current data, and HA switchover, which may interrupt connections. github.com

  • Pre-approved shell tools in the skill

    The frontmatter allows Bash for railway, which, command, npm, npx, curl and python3. It also lists CLI installs through curl piped to a shell, npm and brew. github.com

  • Database scripts connect over SSH

    dal.py runs railway ssh to run queries inside your database service. SKILL.md lists enable-pg-stats.py and pg-extensions.py as commands only you should run. github.com

  • It adds a hosted MCP server

    The plugin's .mcp.json connects to https://mcp.railway.com, authenticated with your Railway account over OAuth. github.com

  • CLI calls carry a caller and session id

    SKILL.md has the agent prefix railway calls with RAILWAY_CALLER=skill:use-railway@1.7.0 and a session id, which it describes as for telemetry. github.com

Technical details
Talks tomcp.railway.com. backboard.railway.com. railway.com. agents.railway.com
Scopes

  • Railway account via browser OAuth or RAILWAY_API_TOKEN / RAILWAY_TOKEN

  • Railway CLI, hosted MCP and GraphQL API

  • Current directory (uploaded on railway up)

Questions

What does the use-railway skill do?

It lets your coding agent deploy to and run things on Railway: projects, services, databases, buckets, variables, domains, logs and metrics. It picks the Railway CLI, the hosted MCP server or the GraphQL API for each task.

Is it official?

Yes. It is in Railway's own railway-skills repository, and the plugin lists Railway as author. Railway also offers a Claude connector that uses the hosted MCP server, covered on our Railway MCP page.

How do I install it in Claude Code?

Use the install box above. The README's Claude Code route is the plugin, which also adds Railway's MCP server and a hook. The skills CLI route installs only the skill.

Do I need the Railway CLI?

For local work like deploying the current directory, SSH and database analysis, yes. The plugin README says the hosted MCP server covers most platform operations without a local install. SKILL.md installs the CLI if it is missing.

Does it work in the Claude app?

The README points Claude app users to Railway's connector, which uses the hosted MCP server and does not need the CLI. That connector is separate from this skill, so we list no skill upload route.

Who pays for what it deploys?

The skill has no price of its own. Services you create live in your Railway workspace, which SKILL.md calls the billing scope. The skill does not state prices, so check Railway's pricing.

Will it delete things without asking?

SKILL.md says to confirm intent and state the impact before destructive actions, and the database-changing scripts are user-only. The plugin's hook, though, auto-approves railway commands, so your client's own prompt may not appear.

Is use-railway safe?

It can deploy your code (railway up uploads the current directory), change variables and delete services using your Railway account, and its plugin includes a hook that auto-approves railway commands. Our advice, not Railway's: try it on a test project first and review commands before they run. See the safety section for the full list.

Information

updated Nov 2, 2026

Repository
railwayapp/railway-skills
Folder
plugins/railway/skills/use-railway

railwayapp/railway-skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 9
  1. github.com · github.com
  2. github.com · github.com
  3. github.com · github.com
  4. github.com · github.com
  5. github.com · github.com
  6. github.com · github.com
  7. github.com · github.com
  8. github.com · github.com
  9. github.com · github.com

Updated Nov 2, 2026.